Should We Use AI In Development?

The short answer is no. In this article I address some of the major trade-offs with using LLMs for development. Things like mental atrophy, lack of ownership, the carbon footprint of AI, and more. As much as AI can be useful for us, it's important to keep these things in mind when we use it. The long answer is that ai usage can range from just insignificant tasks like searching for something or making a small change to scaffolding and perhaps even building full projects. I've used it on almost every stage through my modest journey as a software developer. That increasing reliance on ai blunts the shiny edge of the human brain: Thinking. AI-assistants can save time during the first but whatever it saves the programmer during development will be paid six folds in maintenance. When creating buildings of thoughts, the involvement of the programmer is always an important aspect. Using AI can make you shoot yourself in the foot more often than when you code yourself (and sometimes in much more bizarre ways). Even if you manage to write production-grade, bulletproof code using AI there is the cost of your privacy and the carbon footprint and so on and so forth and all of this assuming that Agents won't accidentally delete your entire production database and lie about it (that actually happened before) We are also hitting a wall with data, since AI can only get so far with what we train it on. That doesn't seem like a totally replacement for engineers, does it? AI is great and can help humanity in countless ways, but I think LLMs should stay away from work where cognitive output (in our case it's code) is the product of the worker. To cement my point, I challenged myself to write this post without any assistance. (not even a spell checker). Let me know what you think about using ai in your field, I'm eager to hear your opinions on the matter.